Could you add LICENSE file?
I plan to make this package, but it need define a license.

ref: kazuhisya/dm-writeboost-rpm/issues/7

Some command takes wbdev and other command takes caching device is confusing to users.

Crates are updated

• clap 3
• crc 2

They are still alpha but when it goes out of alpha stage, we should catch up.

Now this project compiles several binaries but providing them as sub commands is the modern way.

The command name wb may be conflicted with others because it is used in many contexts like "write back" or "write buffer". So the command name should be something like wbtools.

However, the downside of this change will be documents, demos and packages will be all updated. What should I do.

wbcheck -a checks all segments in the caching device

The current implementation returns error if the specified segment is unused yet and uninitialized. The command should return 0 (success) on this case.

Hi again!
I would like to ask what write=1 & hit=1 means?
Example:
cursor pos = 52040677
# of cache blocks = 91033346
# of segments = 716798
current id = 1126568
last flushed id = 1126567
last writeback id = 487778
# of dirty cache blocks = 66562932
# of partial flushes = 38108
write? hit? on_buffer? fullsize?
0 0 0 0 1
0 0 0 1 205034196
0 0 1 0 0
0 0 1 1 0
0 1 0 0 0
0 1 0 1 1681415
0 1 1 0 0
0 1 1 1 0
1 0 0 0 0
1 0 0 1 128200323
1 0 1 0 0
1 0 1 1 0
1 1 0 0 0
1 1 0 1 498264
1 1 1 0 0
1 1 1 1 67

What line 1 1 0 1 498264 means?

Since Rust is evolving the source code should be changed accordingly. Basic regression test is ok but should be written and environment should be built in docker.

akiradeveloper/dm-writeboost#176

I think this code should be fixed too? But user land may submits 512B by RMW manner that results in submitting 4KB request in kernel. I will set the status to bug when reported.

 if id == 0 { // superblock
        let mut buf = [0;512];
        let mut f = File::open(&devname).unwrap();
        f.read(&mut buf).unwrap();
        let sup_header = lib::SuperBlockHeader::from_buf(&buf);

        let s = if sup_header.magic == 0x57427374 {
            "formatted"
        } else {
            "unformatted"
        };

        println!("[superblock header]");
        println!("magic = {} ({})", sup_header.magic, s);

        f.seek(SeekFrom::Start((1u64 << 20) - 512)).unwrap();
        f.read(&mut buf).unwrap();
        let sup_record = lib::SuperBlockRecord::from_buf(&buf);

        println!("[superblock record]");
        println!("last writeback id = {}", sup_record.last_writeback_segment_id);

typical mistake should be avoided. if nr_caches isn't a number but "writeboost" it should be cautioned as error. This occurs when user input the output of dmsetup status instead of dmsetup status wbdev

Writeboost doesn't allow cache device sharing. If user tried to do this, it may result in data corruption.

To avoid this, akiradeveloper/dm-writeboost#115 suggests adding safeguard to the kernel module.

But it may be easier to do with wbcreate command.

reported in kazuhisya/dm-writeboost-rpm#11

clap 3.0 starts to provide with the structopt like parsing.

The first example shows the simplest way to use clap, by defining a struct. If you're familiar with the structopt crate you're in luck, it's the same! (In fact it's the exact same code running under the covers!)

In case the old style parsing is deprecated, better to switch to the new style or use structopt.

Packing everything under a crate isn't the way Rustacean should do.

This argument parser is great!

https://github.com/kbknapp/clap-rs

I will use QuickTime to record the screen.

The video makes it easy to learn how to use this tools and thus dm-writeboost.